Three-plus years building AI that runs in production, not in notebooks.

I've shipped a threat-detection service scoring 500K security events a day, a no-code ML platform that customer teams trained their own models on, and a six-layer medallion architecture on Azure Databricks serving controlled academic access to utility asset data.

Lately my work sits where retrieval, agents, and data engineering meet — grounded RAG with real citations, multi-agent orchestration, and the unglamorous pipeline and evaluation work that decides whether any of it survives contact with users.

Graduate Research Assistant
Prof. Bradley Lehman's group — Dept. of Electrical & Computer Engineering, Northeastern
_{Sep 2026 – present}

Supporting an industry-sponsored research collaboration on utility asset data: building pipelines for asset condition data, de-identification for research use, and applying LLM-based retrieval and evaluation to power systems datasets.

🎓  MS, AI Systems Engineering — Northeastern, Dec 2026
📍  Boston, MA

_{Jan 2026 – Sep 2026 · Marlborough, MA · the industry half of the research collaboration I now support from Northeastern}

• Lead developer on a joint Doble–Northeastern research data platform on Azure Databricks, designing a six-layer medallion architecture with a Unity Catalog hierarchy for controlled academic access to utility asset data
• Built a production RAG system for technical documentation queries — semantic search over vector embeddings returning grounded answers with source citations
• Built a structured extraction tool for dissolved gas analysis reports, combining OCR with schema-driven extraction to process 100+ reports in under six minutes at over 95% accuracy
• Developed multi-agent orchestration workflows in Microsoft Copilot Studio; fine-tuned models with Azure ML Studio and Azure AI Foundry

AI/ML Engineer · Swimlane engagement
_{May 2024 – Aug 2024 · Hyderabad, India}

• Deployed a production ML threat detection service processing over 500,000 security events per day at sub-100ms inference latency
• Built an automated ML pipeline with drift detection and monitoring, improving production prediction accuracy over the prior baseline

AI/ML Engineer · AI Studio
_{Mar 2021 – Jun 2023 · Kochi, India}

• Engineered the core model training capability for AI Studio, a no-code ML platform — TensorFlow, PyTorch and Hugging Face Transformers across classification, detection and NLP
• Automated hyperparameter tuning and distributed training on AWS GPU infrastructure, cutting training time
• Led full-stack development (Python, Django, React) for the AI Studio interface

_{Python CLI GitHub Actions · on PyPI}

Why does your code pass locally and fail in CI? A zero-dependency Python CLI that fingerprints two environments, ranks the likely root cause of the drift, and explains the fix in plain English — with exit codes for use as a CI gate.

Validated against a synthetic fault injector and a real-world corpus: 100% top-1 accuracy across 23 cases, no false positives.

_{TypeScript Pyodide WebAssembly D3 · live demo}

Watch your code run. Executes JavaScript and Python step by step in the browser and visualizes data structures as they mutate — with LeetCode-style practice problems and a built-in judge.

Python runs client-side via Pyodide/WASM, so there's no backend to pay for.

_{Python Node.js Docker · live demo}

An evaluation harness comparing LLMs on text-to-SQL across accuracy, schema comprehension and latency, with per-query scoring.

Fine-tuned models with LoRA on the Gretel AI synthetic text-to-SQL dataset, improving schema grounding and execution accuracy over the base models.

_{PyTorch Stable Diffusion Gradio · Hugging Face Space}

Fine-tuned Stable Diffusion with LoRA using progressive training and memory optimization to fit on a single GPU, paired with a web app for prompt-based generation.

A study in doing real generative work under hard hardware constraints.

Northeastern University — Boston, MA
Master of Science — AI Systems Engineering concentration · _{Sep 2024 – Dec 2026 (expected)}

_{Natural Language Processing · Prompt Engineering and Agentic AI · Knowledge Graphs with GenAI and Databases · Big Data Systems}

Jawaharlal Nehru Technological University — India
Bachelor of Technology · _{Sep 2016 – Jul 2020}

_{Both in progress.}